To give it a definition, an implicit function of x and y is simply any relationship that takes the form:

Transcription

1 2 Implicit function theorems and applications 21 Implicit functions The implicit function theorem is one of the most useful single tools you ll meet this year After a while, it will be second nature to think of this theorem when you want to figure out how a change in variable x affects variable y First a reminder of what an implicit function looks like A non-implicit function (an explicit function?) is the kind of thing you re accustomed to, which looks like: y = f(x) and you call y the dependent variable and x the independent variable But you could also express this same relationship as a zero of the implicit function: g(x, y) = y f(x The pairs of x and y which satisfy the first relationship will also satisfy the second relationship The reason for calling it an implicit function is that it doesn t say outright that y depends on x but it is a function: as you vary x, you have to vary y as well, in order to maintain the equals zero relationship To give it a definition, an implicit function of x and y is simply any relationship that takes the form: g(x, y Any explicit function can be changed into an implicit function using the trick above, just setting g(x, y) = y f(x In theory, any implicit function could be converted into an explicit function by solving for y in terms of x In practice, this may be rather challenging, though Consider: ln(x + y)+ xy 12 = 0 Try as hard as you like, you ll never be able to isolate either of the variables This is one motivation for working with implicit functions Another motivation is that we often work with general functions, rather than a particular functional form For instance, once can generally state the FOC from a typical utility maximization problem in this form: u( v x ) λp l = 0 It would be desirable to talk about properties of demands without assuming that the utility function is Cobb-Douglas, CES, or anything in particular a general form would encompass all these cases Fall 2001 math for economic theory, page 6

2 The final reason to learn how to work with implicit functions is that implicit function naturally arise in economics Every time we do a constrained optimization problem, we end up with some condition set equal to zero This is already an implicit function Why try to solve it for one variable in terms of the others, when we don t need to? These are the sorts of things that we will be asking from the implicit function theorem: Example 211: u w 1 s The first order condition from a utility maximization problem is ( ) + (1 + r) u ( w 2 + (1 + r)s Find ds dr Example 212: The first order conditions from a utility maximization problem are: α αx 1 1 x 1 α 2 λp 1 = 0 (1 α)x α 1 x α 2 λp 2 = 0 Find all the partial derivatives of the demand functions, x i p j 22 The implicit function theorem (two variable case) When we have an implicit function of the form g(x, y, x, y R 1, the implicit function theorem says that we can figure out dy dx quite easily 1 Provided that we have some continuity and a non-zero denominator, this derivative is: g(x, y dy dx = x y Remember that for dy dx, the denominator of the right-hand side is y In other words, whatever variable is on top on the left is the derivative which is on bottom on the right One trick to remember this is to pretend that you can cancel out the whatever terms, so that you get: x y = g / x g / y = 1 x 1 y = y x This is not how the implicit function theorem works, though It s just a useful trick for checking what goes where 1 At least, you can figure it out quite easily provided that you know which derivative you re trying to find these variables aren t going to be called x and y in the problems you encounter, they ll be s and r and p and h and w Half the time, the question won t ask, Find s r as in Example 211 instead it ll be worded, Find how much savings changes when the interest rate changes It ll be up to you to figure out what that means Fall 2001 math for economic theory, page 7

3 So how does the implicit function theorem work? Again, an implicit function is like a normal function in that y must change as x changes, in order to maintain the relationship You can think of y(x) The function is really just a function of one independent variable, like this: ( g x, y(x) There s a y in there, but not really the value of y is dictated by x By totally differentiating this function with respect to x, we get: d dx [ g( x, y(x) )] = d [ dx 0 ] x + y y (x) = 0 dy dx = y (x) = x y And thus, we get the implicit function theorem by doing nothing more than treating y as a function of x and totally differentiating An alternative way of deriving this result is to just take the function g(x, y), and write its change in differential form Since we know that 0 is unchanging, this form is: dy + y x dx = 0 dy = y x dx dy x = dx y The derivation of the implicit function theorem is quick and simple, and it might be work remembering in case you can t remember the theorem itself 23 ultivariate versions of the implicit function theorem When y is an implicit function of many variables Now we re talking about implicit functions that look like: g, y It turns out that nothing really changes for these functions If we write out the differential form of this function, we get: dx 1 + dx 2 +K + dx n + x 2 x n y dy = 0 Fall 2001 math for economic theory, page 8

4 By setting all the dx i = 0 except for one dx l, we get the partial derivative of y with respect to x l This expression turns out to be much the same as for the single-x case: y = y When there are many y variables and many x variables Let s say that we have m variables that we call y 1,K y m, and that these variables are implicitly a function of some n variables, called x 1 It will take m equations to describe the entire system of y variables, like: g 1 g 2 g m There are several ways to express the implicit function theorem in this form One is to imagine that those m implicit functions are a single vector-valued function g:r n m R m, such that: g( v x, v y v If we write this in differential form, we get that: Dx v ( x v, v y ) dx { + D v y ( v x, y v ) n dy { = 0 { m 1 m n m m m 1 (You can verify that the dimensions make sense, right?) You can sort of solve this equation to get one multivariate formulation of the theorem: dy = D v y g( x v, v y ) 1 Dx v g( v x, v y ) dx Alternatively, you could think about only the partial derivatives of y with respect to one of the x variables To find this out, you do: T,,K = Dy x i x i x i v g( v x, y v ) 1 D xi g( v x, v y ) Finally, we can use a version of Cramer s rule to solve systems of implicit functions Suppose that we re given: g 1 g 2 g m and we want to know one partial derivative, like y k this function looks like: The differential form of Fall 2001 math for economic theory, page 9

5 1 dy 1 +K + 1 dy m + 1 dx 1 + K+ 1 x n dx n = 0 m dy 1 +K + m dy m + m dx 1 +K + m x n dx n = 0 Now, the idea of a partial derivative is that only x l changes and none of the other x variables; but this can still mean that lots of y variables change around, not just y k So we set dx i = 0 for all i l (which means that we re really talking about a partial derivative at this point holding everything constant except dx l I ll change the notation to partial derivatives) This results in a system of m equations in m unknowns these unknowns are the y j : 1 +K + 1 = 1 m +K + m = m (See how it s a system of linear equations?) There are several ways that we can solve this system of equations one of them being Cramer s rule Recall that if we want to solve for the unknown variable y k on the left-hand side, we establish the m m matrix of coefficients on the unknown y j variables this turns out to be the same as the matrix of first derivatives, D y g( v x, y v ) Cramer s rule says that y k is equal to a fraction, the denominator of which is det( D y g( v x, y v )), the numerator of which is the determinant of the same matrix with the k-th column replaced by the vector on the right-hand side, or ore clearly: 1 L 1 L 1 det m y k L m L m = l L 1 L 1 det m y1 L m L m That s the formula to remember for the multivariate version of the implicit function theorem Note that the minus sign in front of the in the system of equations above has turned into a minus sign out in front of the fraction This comes because we ve changed the signs of every element in a row of the matrix in the numerator; there s a rule which says that multiplying every element of one row or column of a matrix by some scalar α means that the determinant of the matrix also changes by the same α In this case, α = 1 Fall 2001 math for economic theory, page 10

6 24 Applications and exercises Some applications of the implicit function theorem show up in micro, but most of them are in macro Here are some problems that you will face: Exercise 241: For the utility function u, x 2 ), a particular vector, x 2 ) will put you on some indifference curve u Find the slope of this curve at, x 2 ) Step 1: Figure out what derivative you want Step 2:Define an implicit function of these variables Step 3: Apply the implicit function theorem Exercise 242: The individual lives for two periods He has a utility function u(c 1 )+ βu(c 2 ) His budget constraint requires that his period 1 consumption be his period 1 endowment minus any savings, c 1 = w 1 s In the second period, his consumption will be c 2 = w 2 + (1 + r)θs + α The government has taxed savings at the rate 1 θ, and uses it to finance a lump-sum transfer of α The individual takes θ and α as given, but for the government the government s budget to balance, we must have α = (1 θ)(1 + r)s Find s θ Step 1: Set up the utility maximization problem Step 2:Find FOCs These should define an implicit function Step 3: Insert the GBC, after taking the FOC Step 4:Apply the implicit function theorem Exercise 243: People live for two periods (in overlapping generations), and the utility function is the same as before The government has implemented a pay-as-you-go social security system The way this works is that every young person pays a lump-sum tax of θ, and every old person collects a pension of α This means that c 1 = w 1 θ and c 2 = w 2 + (1 + r)s + α The size of each cohort (or generation) increases at rate n, which means that N t +1 = (1 + n)n t, where N t represents the number of people born at time t Find s θ Step 1: Set up the utility maximization problem Step 2:Figure out what the GBC is supposed to be Step 3: Find FOCs Step 4: Insert the GBC Step 5: Apply the implicit function theorem Fall 2001 math for economic theory, page 11

7 Exercise 244: The individual lives for two periods His utility function depends on consumption at time 1 and consumption at time 2, u(c 1,c 2 ) He is endowed with w 1 when young and w 2 when old He can chose savings s at t = 1, which get a gross rate of return 1 + r For each of the following utility functions, find s r a u(c 1,c 2 ) = ln c 1 + β ln c 2 b u(c 1,c 2 ) = ln c 1 + β c 2 c u(c 1,c 2 ) = (1 ρ) 1 c 1 1 ρ + β c c 1 ρ { }, and w 2 = 0 d u(c 1,c 2 ) = exp( ρc 1 )+ exp( ρc 2 ), and w 2 = 0 Step 1: Sep up the utility maximization problem Step 2:Find FOCs Step 3: Evaluate: is it easier to solve for s, or to use the IFT? Step 4:Do it If you have some extra time, you might want to try solving those problems using both methods solving for s directly, as well as using the implicit function theorem Which way is easier? What s different about the way the answers look? Can you reconcile them? Exercise 245: The individual has a utility function represented by u, x 2 ) = x α 1 α 1 x 2 Using the first order conditions, find x i p j Now solve the demand functions, and take the same derivatives Do they match up? Why or why not? Exercise 246: Same thing, except that the utility function is now ( ) 1/ρ u, x 2 ) = x 1 ρ + x 2 ρ Fall 2001 math for economic theory, page 12